使用 Swift 的 Microsoft 认知语音连续识别

Question

我正在尝试使用适用于 iOS 的 Microsoft Cognitive Services Speech to Text SDK。到目前为止，我已经设法让识别Once 函数运行没有任何问题，但是 Swift 的样本很少，而且我似乎无法使用 Swift 进行连续识别。请记住，我通常使用 React Native/JS 并且对 Swift 有非常基本的了解。话虽如此，任何帮助将不胜感激。这是我目前所拥有的：

func recognizeFromMic() {

    var speechConfig: SPXSpeechConfiguration?
    do {
        try speechConfig = SPXSpeechConfiguration(subscription: key, region: region)
    } catch {
        print("error \(error) happened")
        speechConfig = nil
    }
    speechConfig?.speechRecognitionLanguage = "en-US"

    let audioConfig = SPXAudioConfiguration(microphone: nil)

    let reco = try! SPXSpeechRecognizer(speechConfiguration: speechConfig!, audioConfiguration: audioConfig!)

    reco.addRecognizingEventHandler() {reco, evt in
        print("Received intermediate result event. SessionId: \(evt.sessionId), Recognition Result: \(evt.result.text ?? "(no result)"), Status: \(evt.result.reason), Offset: \(evt.result.offset) Duration: \(evt.result.duration), ResultID: \(evt.result.resultId)")
    }

    reco.addRecognizedEventHandler() {reco, evt in
        print("Received final result event. SessionId: \(evt.sessionId), Recognition Result: \(evt.result.text ?? "(no result)"), Status: \(evt.result.reason), Offset: \(evt.result.offset) Duration: \(evt.result.duration), ResultID: \(evt.result.resultId)")
    }

    reco.addSessionStoppedEventHandler() {reco, evt in
        print("Received session stopped event. SessionId: \(evt.sessionId)")
    }

    print("Listening...")

    do {

        try reco.startContinuousRecognition()

    } catch {

        print("error \(error) happened")

    }
}

运行上述函数时出现以下错误：

Error Domain=SPXErrorDomain Code=21 "Error: Exception with an error code: 0x15 (SPXERR_MIC_ERROR)

Answer 1

因此，经过巨大的努力，我终于有了一个工作功能。微软绝对应该改进他们的例子。对于那些寻找相同答案的人来说，这就是我最终要做的。希望这对某人有帮助:)

var audioEngine = AVAudioEngine()
var audioData = Data()
var inputNode: AVAudioInputNode?
var reco: SPXSpeechRecognizer?;
var stream = SPXPullAudioInputStream()

func recognizeFromMic() {

    var bytesPerFrame = UInt32()
    var stream: SPXPushAudioInputStream?

    func toData(PCMBuffer: AVAudioPCMBuffer) -> Data {
        let channelCount = 1  // given PCMBuffer channel count is 1
        let channels = UnsafeBufferPointer(start: PCMBuffer.int16ChannelData, count: channelCount)
        let ch0Data = NSData(bytes: channels[0], length:Int(PCMBuffer.frameCapacity * PCMBuffer.format.streamDescription.pointee.mBytesPerFrame)) as Data
        return ch0Data
    }

    guard let outputFormat = AVAudioFormat(commonFormat: .pcmFormatInt16, sampleRate: 16000, channels: AVAudioChannelCount(1), interleaved: false) else {
        return
    }

    let audioSession = AVAudioSession.sharedInstance()
    do {
        try audioSession.setCategory(.record, mode: .measurement, options: .duckOthers)
        try audioSession.setActive(true, options: .notifyOthersOnDeactivation)
    } catch {

    }
    inputNode = self.audioEngine.inputNode
    let inputFormat = inputNode!.outputFormat(forBus: 0)
    let converter = AVAudioConverter(from: inputFormat, to: outputFormat)
    inputNode!.installTap(onBus: AVAudioNodeBus(0), bufferSize: AVAudioFrameCount(32000), format: self.audioEngine.inputNode.outputFormat(forBus: 0)) { (pcmBuffer, time) in

        guard let outputBuffer = AVAudioPCMBuffer(pcmFormat: outputFormat, frameCapacity:pcmBuffer.frameCapacity) else {
            return
        }
        outputBuffer.frameLength = pcmBuffer.frameLength
        let inputBlock: AVAudioConverterInputBlock = { inNumPackets, outStatus in
            outStatus.pointee = AVAudioConverterInputStatus.haveData
            return pcmBuffer
        }
        var  error:NSError? = nil
        converter?.convert(to: outputBuffer, error: &error, withInputFrom: inputBlock)
        self.audioData.append(toData(PCMBuffer: outputBuffer))
        stream?.write(self.audioData)

    }
    self.audioEngine.prepare()
    do {
        try self.audioEngine.start()
    } catch {
        print("\(error)")
    }

    //set up push stream
    let audioFormat = SPXAudioStreamFormat.init(usingPCMWithSampleRate: 16000, bitsPerSample: 16, channels: 1)
    stream = SPXPushAudioInputStream(audioFormat: audioFormat!)

    var speechConfig: SPXSpeechConfiguration?
    do {
        try speechConfig = SPXSpeechConfiguration(subscription: key, region: region)
        speechConfig!.enableDictation();
    } catch {
        print("error \(error) happened")
        speechConfig = nil
    }
    speechConfig?.speechRecognitionLanguage = "en-US"

    let audioConfig = SPXAudioConfiguration(streamInput: stream!)

    reco = try! SPXSpeechRecognizer(speechConfiguration: speechConfig!, audioConfiguration: audioConfig!)

    reco!.addRecognizingEventHandler() {reco, evt in
        print("Received intermediate result event. SessionId: \(evt.sessionId), Recognition Result: \(evt.result.text ?? "(no result)"), Status: \(evt.result.reason), Offset: \(evt.result.offset) Duration: \(evt.result.duration), ResultID: \(evt.result.resultId)")
    }

    reco!.addRecognizedEventHandler() {reco, evt in
        print("Received final result event. SessionId: \(evt.sessionId), Recognition Result: \(evt.result.text ?? "(no result)"), Status: \(evt.result.reason), Offset: \(evt.result.offset) Duration: \(evt.result.duration), ResultID: \(evt.result.resultId)")
    }

    reco!.addSessionStoppedEventHandler() {reco, evt in
        print("Received session stopped event. SessionId: \(evt.sessionId)")
    }

    print("Listening...")

    // Start recording and append recording buffer to speech recognizer

    do {
        try reco!.startContinuousRecognition()
    } catch {
        print("error \(error) happened")
    }

}